The Future of Humanity- Yuval Noah Harari
What I want to discuss today is one particularly important possibilities, the possibilities that we are now facing, which is the rise of artificial intelligence, and the potential that artificial intelligence will become the dominant life form on earth, and even beyond the earth. And what would that mean for Homo sapiens. And in order to understand this, we need to maybe go a step backward, and take the really long view, not just of history, but actually of biology, not just of the history of men, but really of the history of life. Because what’s happening now is maybe not just the most important revolution in human history, but the most important revolution in biology since the beginning of life, at least on Earth.
For the last 4 billion years of Earth, sorry of life, nothing much changed about the fundamental laws of life. all life forms for 4 billion years evolved by natural selection. And all life forms were restricted to the organic real. It doesn’t matter if you’re an amoeba, or a giraffe or a tomato or a homo sapiens, you’re made of organic compounds and your subjects, you are subject to organic biochemistry. This is now changing. We are on the verge not only replacing natural selection with intelligent design, not the intelligent design of some God. Our intelligent design is the principal motor of the evolution of life. We are also on the verge of allowing life to break out for the first time from the organic realm into the inorganic, and creating the first inorganic
life forms after 4 billion years of evolution, which also implies that life will for the first time have a serious chance of breaking out of planet Earth, and beginning to spread in the rest of the galaxy in the rest of the universe. organic life has been adapted to the very unique conditions of this planet for 4 billion years. So despite what you see on Star Trek, which is just now celebrating 50 years, it’s extremely difficult to sustain organic life in outer space on other planets, which is why it’s very unlikely that organic life will be able to spread them to flourish outside Earth. But once you switch from organic to inorganic, it becomes much easier. So we are really on the verge of these two intertwined revolutions, moving from the organic to the inorganic and moving
from Planet Earth, actually to the rest of the universe to the rest of the galaxy.
But coming back from the level of the galaxy, to the level of day to day politics, what this what will this mean for ordinary human beings for society for the job market for the political system, I don’t have time to cover all the different scenarios and possibilities, I want to focus on one very important issue, which is what it will do to the job market and what it will do to the economic
importance and power of ordinary humans in the next few decades. Because this is a very practical question.
Kids in the UK have just started the school year this week. And the question is, what do they need to learn so that they will still be relevant, they will still have a job when they are 30 or 40.
This is it this kind of question, which I think, brings down the idea of artificial intelligence from the cloud and the galaxy to the level of society and economics and politics.
And one of the big dangers, which more and more experts are talking about is that in the next few decades, not in thousands of years, but within the lifetime of many of the people here in the room, artificial intelligence will push humans out of the job market. And in the same way, that the industrial revolution of the 19th century created a new massive class, the urban working class, the proletariat. So in the 21st century, a new industrial revolution will create a new massive class the useless class people who have no economic usefulness because AI artificial intelligence outperforms them in almost all times.
And old jobs, people who are not just unemployed, they’re unemployable. There is no, there is no jobs, no jobs to give these people. So to start with a simple example 10 years ago, it was relatively accepted wisdom that computers and artificial intelligence will never be able to drive cars and vehicles better than humans may be in a laboratory. Under sterile conditions. A computer could drive a car, but not in real-life situations in a real city like London. Today, more and more experts are coming to the opposite conclusion to the opposite view. But it’s only a question of time and not a very long time. Maybe in 10 2030 years, humans will not drive vehicles at all, because AI will be so much better at driving taxis and buses and trucks and so forth.
Human beings, they will not only drive it more efficiently and more cheaply with less pollution. But most importantly, AI will drive far more safely than homosapiens than human drivers. Today in the world, every year 1.3 million people die each year from car accidents around the world. This is about twice as many deaths as the death caused by war, crime and terrorism put together. And most of these 1.3 million people are killed as a result of human mistakes of human arrows or negligence. It’s cases like people falling asleep while driving cases like people drinking alcohol and then driving people ignoring traffic lights, people ignoring the stop sign, people driving too fast inside a city, things like that. This is the major
reason for this more than 1 million deaths every year. If you replace humans, by AI, most of these arrows will disappear will be prevented. And self-driving car a computer will never drink alcohol and then drive.
A self-driving car will never fall asleep while driving. If you program it correctly, or if this is how you program it, it will never disobey the traffic laws, it will always stop at a stop sign, it will always stop at a red sign. Even more importantly, from the viewpoint of safety.
Today, each car is an individual unit. When two cars are approaching an intersection. Each car sometimes tries to signal it intention the driver is trying to signal his or her intentions, but the two counts are really independent entities which
why sometimes they collide. However, if you prevent humans from driving, and all the vehicles on the road, I’ll sell our autonomous and they are driven by computers by AI, then the logical and possible thing to do is to connect all of them want to the others, so they are no longer independent vehicles on the road. But all vehicles are connected to a single network to a single master algorithm, which is far less likely to allow two of its puppets to collide.
So from this perspective, it’s quite likely it’s not it’s not a profit. It’s not certain all kinds of things may prevent it from happening. But there is a good chance that insane 10 2030 years humans all human drivers are most human drivers will be replaced by AI, and which we should have enormous good consequences.
Which also means that millions of jobs will disappear.
The same thing may happen with many, many other professions. I’ll just I don’t have a lot of time. So I’ll give just one example. what might happen to doctors?
Most doctors, what they do most of the time is to try to diagnose a disease and then offer the best treatment possible, which is something I’m very aware of right now, because I’m a bit under the weather. I drove to Manchester yesterday and go to cold on the way. I think it’s a cold, I’m not sure. I feel a bit dizzy. I feel a bit something in the throat. I think it’s a cold. I can’t go and ask my doctor because my doctor is back in Israel. And even if I could, if I even if I was right now in Israel, it’s not so simple to go to the doctor. It takes time. I need to make an appointment.
She’s not always available. And even if I make an appointment for tomorrow morning, so I have to leave my work and drive to the clinic, and then wait in the reception room for 10 minutes or 20 minutes or 30 minutes. And then finally I get to see the doctor. And I don’t know how it is with the NHS in the UK. But in Israel, my insurance pays for very short visits, maybe five or 10 minutes. That’s all that my doctor usually has for me during these five or 10 minutes when she tries to diagnose my disease, so she would ask me three four questions about how I’m feeling. Do you do you have a headache? Do you feel dizzy, something like that? She may do one or two simple physical tests. She may ask me to say I look into my throat. She may take out a stethoscope and listen to my lungs or my heart. She may measure my heartbeat to my blood pressure. She also knows something about my medical history, because she’s my personal physician.
But obviously, she can’t remember every illness I ever had. And every blood test and DNA scan I are ever made, she may look it up on the computer. But again, she doesn’t have much time. So she takes these few bits of data about my present and past medical condition. And now, in order to diagnose my disease, she needs to compare that with all the different diseases in the world, could be called could be influenced, could be a breast cancer, all kinds of things that might have these symptoms. And obviously, even the best doctor in the world, she can’t really be familiar with all the different medical conditions and all the different diseases in the world. And even if, and she obviously also, she can’t be updated every day about all the latest medical researchers and articles and tests and drugs and so forth. So
Both types of data, what she knows about me and what she knows about all the diseases and medical conditions in the world, both are very limited. In addition, my doctor is sometimes sick herself. She sometimes irritated, she is sometimes hungry, she sometimes tired, so she doesn’t always at her peak of her performance when she comes to diagnose my disease. Now compare that to AI doctors, artificial intelligence doctors that are already being developed as we speak. The most famous example but not the only example is IBM Watson.
Watson has immense advantages compared to my flesh and blood physician. First of all, Watson can be everywhere all the time on my smartphone. Even if I go to give a talk him in London, I can take my personal physician with me on the smartphone
Companies need 24 hours a day, 365 days a year.
It has all the time in the world for me. If I want I can sit on my living room so sofa and just answer questions about my health for hours and hours on end and do all kinds of tests.
In fact, Watson doesn’t need to wait until I asked until I come to Watson. They say hey, something is wrong, what’s wrong with me? Watson will be able to monitor my medical condition all the time using biometric sensors on my body and inside my body. So when something just starts, it will know about it much month before I know that something is wrong, and it can try to do something about it to start the treatment even without my knowing it.
In addition, what’s on has no limited or almost no limitations on the amount of data, it can access and process what
We’ll be able instantly to know my entire medical history, every illness I ever had every blood test or every DNA test I ever did. In addition, what some will be able to access such data about my parents and siblings and neighbors and friends and then strangers.
The other side of the equation, what about all the diseases in the world he to Watson has immense advantages over flesh and blood doctors, Watson will be able, to a large extent to be familiar with all the different diseases in the world, and with all the newest latest medical research, about disease, about drugs about these treatments, that treatment. So from this perspective, it’s very likely that Watson will be able to diagnose disease and to offer treatment far far better than any human doctor.
When people hear this, they very often say okay, maybe Watson will be better in diagnosing disease. But there is one thing one other thing that we usually hope human doctors will do and that Watson will not be able to do and this is offer emotional support. A human doctor is not some machine some cold machine that just diagnosis disease and says take this pill, a good doctor is also very attentive to my emotional condition. And it not just treats my physical difficulties, it also gives me the proper emotional support that in many cases is a vital part of a of confronting any kind of disease or medical condition. However, this criticism fails to, to know to noticed that emotions, at least according to modern science, emotions are not some spiritual thing. Good
That God gave humans in order to appreciate poetry. Emotions are a biochemical phenomenon that not only Homo sapiens have all mammals, old birds and many, many other animals have emotions that are a biochemical phenomena. In this sense, emotions are like disease, they are both biochemical phenomena. And therefore, it is extremely likely that Watson will be able to diagnose my emotional condition just as a diagnosis, my illnesses and my medical problems.
If I go to my coma, human Doctor, how does my human doctor know my emotional condition? She relies on two kinds of signals, external signals that I’m giving. She relies basically on visual signals, like my facial expression, or my body movement, my body language, and she relies on audio signals, audio cues, she listens.
to what I say, not just the content, but even more importantly, the tone of voice. So if I sit in an office, she looks at my face, she listens to my words, and this is how she knows if I’m angry, if I’m fearful or whatever,
what some will be able to do all that. Computers are already outperforming humans in a diagnosing correctly analyzing correctly, facial expressions, and tone of voice in order to recognize emotions, but much more importantly, an AI like Watson will have access to another and even better source of data about my emotions, data coming from within my body. When my doctor looks at me and I sit in her office, and she looks at me, she sees my face, but she can’t see my brain and she can’t see my heart and she can’t see what’s happening inside me. Watson will be able to access by
metric data coming from the brain coming from the heart coming from the bloodstream, and therefore is likely to be able to diagnose my emotional condition Far, far better than any human doctor.
Now, there are still problems, some technical problems and also legal problems that prevent what’s on and things like Watson, from replacing most doctors tomorrow morning to may take five years, 10 years, 20 years. But what we need to realize that we need to solve these technical problems just once just once. In the case of human doctors of flesh and blood doctors. In order to get to get a doctor, you need to take a person and then you need 10 years at least 10 years of going to medical school and doing all kinds of studying and in experimenting and experiencing in order after 10 years and a huge investment.
And in time and money and energy. At the end of this process, you get one doctor. If you want another doctor, you have to start all over again and invest all this time, all this energy all this money again, which is why in many countries around the world, there is an acute shortage of doctors.
With Watson with an AI doctor, you just have to do it once, even if it costs 100 billion dollars to solve the technical problems that still prevent Watson from replacing my human doctor. If you invest these hundred billion dollars and solve the problems, what you get is not one doctor, you get an infinite number of doctors available everywhere, all the time for everybody. Even somebody in the middle of the jungle can have a personal physician on his smartphone or her smartphone, which for
Far better medical care than almost any doctor alive today. So the potential is really is really immense.
Which is why again, more and more experts believe that not all doctors, but many doctors, maybe 5060 80% of doctors will be replaced by AI within 10 2030 years. And the same thing may happen to many other professions, lawyers, teachers,
insurance agents, and so forth.
When people hear about this possibility, again, one of the most common objections is to say we’ve heard it before. We’ve heard it before this fear of machines replacing humans. It’s not it’s not new. A lot of people in the 19th and 20th century were afraid that as machines replace humans in other
Agriculture and then in industry, you’ll have this massive unemployment and massive crisis with all the useless people. And it didn’t happen, because as new jobs disappeared, sorry, as all jobs disappeared, new jobs appeared to replace them. So we don’t have a crisis today of mass unemployment. What happened is that most people in advanced societies stopped working in the agricultural industry. Today in a country like the UK or the USA, about 2% of the workforce are employed in agriculture, compared to more than 90% before the Industrial Revolution, maybe 15% 20% still working industry physical jobs, the vast majority are working in services. But the problem is, we cannot be sure that the same thing will happen again, with this new revolution. Because humans have basically two kinds
abilities, they have physical abilities, and they have mental and cognitive abilities. What’s happened in the 19th and 20th century is that machines competed and outperform humans in physical abilities. So humans mostly moved to working in jobs that require mental and cognitive abilities like the services sector. Now, machines are starting to compete with us and outperform us also in mental and cognitive abilities. And we just don’t know about any third kind of ability that humans may have, and that everybody could move to work in that field. We just not know about it. Another problem is that even if new jobs appear,
the pace of change is so quick that humans will have to reinvent themselves again and again during their lifetime, which is something that is very very difficult because
beyond a certain age, when you’re 15 of your 20, the main thing you do in life is basically to to invent yourself or to reinvent yourself. And even then it’s not very easy. But when you’re 40, or 50, it’s much, much more difficult. And let’s say that in 20 years, don’t know jobs for taxi drivers and doctors and insurance agents, and they have to reinvent themselves is, let’s say designers of virtual worlds. Now, this is something very difficult for a 50-year-old taxi driver or a 50-year-old insurance agent to invent himself or herself again, as a completely new kind of person. Now I am the creator of virtual worlds. How does the virtual world created by a 50-year-old insurance agent look like?
It’s a very difficult question. And it goes back to what I started with, which is what to do.
teach children today at school? And the answer is that nobody has a clue. I mean, children are today’s star the first week of school first grade, nobody knows what the job market would be like, when these kids are 30 or 40. It’s very likely that almost everything they learn at school will be completely irrelevant. But what to teach them instead, nobody knows. Because we just don’t know what kind of job market will have in 2015, and what kind of skills people will need in these kinds of so far unknown jobs.
as I said, just as the 19th century created this new massive class, the urban working class, and much of the political and social history of the 19th and 20th century revolved around this new class, the working class
Similarly, in the 21st century, we may see the creation, the rise of a new massive class, the useless class, and much of the social and political and economic history of the 21st century will revolve around that class and around the question what to do with billions and billions of economically useless people. And because they are economically useless, the danger is also that they will be politically powerless, because usually, economic usefulness goes hand in hand with political power. And as humans lose their economic usefulness, they may also lose their political power. Going to give just one simple example
when you have all these millions of taxi drivers and truck drivers and bus drivers, each of them commands a small share of the transportation market and this gives them not only
certain amount of economic wealth, it goes to gives them a certain amount of political power, they can unionize. And if the government pursues a policy, which all these taxi drivers or truck drivers don’t like, they can go on strike. So they have also some political power. Now, if you replace all these drivers with driverless cars, which are basically managed by a single algorithm, which is owned by a single Corporation, which is owned by a handful of billionaires. So all the economic and political power that was previously shared between millions of drivers, is now being monopolized by maybe five or 10 individuals.
As I said, in the beginning, this is not a prophecy. Nobody really knows how the job market how the economy how the political system would look like in 20.
50 if you don’t like this particular possibility, you can still do something about it. Thank you